基于形态学脑网络连接模型预测脑小血管病患者的认知功能

Prediction of cognitive function in patients with cerebral small vesseldisease based on morphological brain network connection model

目的 构建脑小血管病(cerebral small vessel disease, CSVD)患者形态学脑网络并预测其认知功能。方法 回顾性选择2020年1月至2024年2月南京医科大学附属江宁医院神经内科住院的老年CSVD患者64例。完善临床资料收集、认知功能评估、多模态磁共振成像扫描等。认知功能评估包括简易智能状态检查量表(mini-mental state examination, MMSE)评分和蒙特利尔认知评估量表(Montreal cognitive assessment, MoCA)评分。利用3D T_(1)加权成像基于Kullback-Leibler散度的相似性方法构建个体形态学脑网络,并结合连接组预测模型方法构建认知预测模型。结果 与MMSE评分和MoCA评分呈显著正相关的网络主要位于默认网络内,且利用显著正相关的形态学脑网络可有效预测个体MMSE评分和MoCA评分(r=0.795,P=4.436×10^(-15);r=0.794,P=4.974×10^(-15),P<0.01)。与MMSE评分和MoCA评分呈显著负相关的连接主要位于凸显/腹侧注意网络与其他网络之间,也可有效地预测个体MMSE评分和MoCA评分(r=0.766,P=1.679×10^(-13);r=0.850,P=6.915×10^(-19),P<0.01)。联合正相关与负相关连接网络,模型预测能力进一步提升(r=0.849,P=7.603×10^(-19);r=0.888,P=1.445×10^(-22),P<0.01)。结论 基于个体形态学脑网络可有效预测CSVD患者认知功能评分,可以作为早期预警CSVD相关认知障碍的一种便捷工具。

Objective To construct a morphological brain network in patients with cerebral smallvessel disease(CSVD)and predict it application for cognitive function.Methods A total of 64 eld-erly CSVD patients admitted in our hospital from January 2020 to February 2024 were retrospec-tively recruited.Cognitive function was assessed with Mini-Mental State Examination(MMSE)and Montreal Cognitive Assessment(MoCA).Their clinical data,and results of cognitive functionand multi-modal MRI scanning were collected and analyzed.3D Ti-weighted imaging based on Kullback-Leibler divergence similarity was used to construct individual morphological brain net-work,and the connectome-based predictive model was employed to construct a cognitive prediction model.Results The network,which is significantly and positively correlated with the MMSE and MoCA scores,was mainly located in the default mode network,and could effectively predictindividual MMSE and MoCA scores(r=0.795,P=4.436×10^(-15) r=0.794,P=4.974×10^(-15) P<0.01).The connections,which were significantly negatively correlated with MMSE or MoCAscores,were mainly located between the salience/ventral attention network and other networks,and could also effectively predict individual MMSE and MoCA scores(r=0.766,P=1.679×10^(-13);r=0.850,P=6.915×10^(-18),P<0.01).Combined positive correlation and negative correlation networks,the model showed further improved predictive performance(r=0.849,P=7.603×10^(-19),r=0.888,P=1.445×10^(-22),P<0.01).Conclusion Individual morphological brain networkcan effectively predict cognitive function in elderly CSVD patients,and can be used as a convenienttool for early warning of cognitive impairment related to CSVD.